Term 3 · Module 4 of 4

Multiple Linear Regression II

Business Statistics for Entrepreneurs III

Practise this module in Claude, ChatGPT or Gemini

Set it up free
Mind map

Categorical Predictors in Multiple Linear Regression

Regression can use qualitative information such as brand, location, ownership, or smoking history alongside numerical measurements. A categorical predictor lets the model represent group differences while retaining quantitative predictors such as price or age.

For a response YY, quantitative predictor X1X_1, and binary indicator X2X_2,

Yi=β0+β1X1i+β2X2i+εi,X2∈{0,1}.Y_i=\beta_0+\beta_1X_{1i}+\beta_2X_{2i}+\varepsilon_i,\qquad X_2\in\{0,1\}.

The fitted coefficients are obtained by least squares as in ordinary multiple regression. Model assessment also follows the same pattern: use R2R^2 and adjusted R2R^2, the overall FF-test, individual tt-tests, and residual diagnostics. The indicator simply adds a predictor column to the analysis.

Categorical Predictors in Multiple Linear Regression

Binary indicators and a computer-brand example

An indicator (dummy) variable codes group membership as 11 for the named group and 00 for its comparison group. In a customer-satisfaction data set with 36 respondents, YY is satisfaction score, X1X_1 is computer price in lakhs, and X2=1X_2=1 for Lenovo and 00 for Dell.

The price-only model had R2=0.36R^2=0.36, F=18.9F=18.9, and a price coefficient t=4.36t=4.36, with p-values close to zero. Adding brand yielded R2=0.45R^2=0.45, standard error 4.854.85, SSR=648.2SSR=648.2, SSE=777.4SSE=777.4, and unchanged SST=1425.6SST=1425.6. The overall FF-statistic was 13.7513.75 with p-value close to zero, rejecting the joint null that both slope coefficients are zero.

The fitted coefficients were b0=46.83b_0=46.83, b1=1.29b_1=1.29, and b2=3.93b_2=3.93. For price, (t=b_1/SEb1b_1=1.29/0.283\approx4.58), with 33 degrees of freedom and p-value close to zero. For brand, (t=b_2/SEb2b_2=3.93/1.63\approx2.42), also with 33 degrees of freedom; the source reports its p-value as close to zero. Thus both predictors were treated as significant in this analysis.

Reading coefficients as group comparisons

The indicator coefficient compares the named group with the reference group at the same values of the other predictors. For Dell (X2=0X_2=0) and Lenovo (X2=1X_2=1), the population mean equations are:

E[Y∣Dell,X1]=β0+β1X1E[Y\mid\text{Dell},X_1]=\beta_0+\beta_1X_1 E[Y∣Lenovo,X1]=β0+β2+β1X1.E[Y\mid\text{Lenovo},X_1]=\beta_0+\beta_2+\beta_1X_1.

The two groups have a common slope β1\beta_1, while their intercepts differ by β2\beta_2. A positive β2\beta_2 means a higher mean response for Lenovo than Dell at a fixed price; a negative value means lower; zero means no modeled difference.

Using the estimates, the predicted equations are approximately:

Y^Dell=46.83+1.29X1\widehat{Y}_{\text{Dell}}=46.83+1.29X_1 Y^Lenovo=(46.83+3.93)+1.29X1=50.76+1.29X1.\widehat{Y}_{\text{Lenovo}}=(46.83+3.93)+1.29X_1=50.76+1.29X_1.

Thus the estimated mean Lenovo satisfaction score is 3.933.93 points above Dell at the same price. (The intercept is also rounded to 50.7750.77.) Residuals should be inspected against each predictor and fitted values. Binary predictors naturally produce two vertical clusters; look for positive and negative residuals without a systematic pattern. Standardized residuals beyond roughly ±2\pm2 merit attention, not automatic deletion.

More than two categories: k−1k-1 indicators

For a categorical variable with kk levels, use k−1k-1 binary indicators and leave one level as the reference (baseline) category. Coding every level with its own indicator alongside an intercept creates redundancy; coding a category as 0,1,20,1,2 incorrectly imposes equal numerical spacing and a linear trend across labels.

For brands Lenovo, Dell, and Asus, choose Asus as baseline and define X1=1X_1=1 for Lenovo and X2=1X_2=1 for Dell (each zero otherwise):

Brand X1X_1 Lenovo X2X_2 Dell Expected score
Asus (baseline) 0 0 β0\beta_0
Lenovo 1 0 β0+β1\beta_0+\beta_1
Dell 0 1 β0+β2\beta_0+\beta_2

Here β0\beta_0 is the expected response for Asus, β1\beta_1 is Lenovo minus Asus, and β2\beta_2 is Dell minus Asus. The baseline choice is arbitrary: choosing Lenovo instead changes the coefficient comparisons and intercept, but not the fitted group means or model predictions.

Key takeaways

  • Code a two-group category with one (0/1) indicator; its coefficient is an adjusted group difference.
  • With (k) categories, use k−1k-1 indicators and interpret each against the chosen baseline.
  • Categorical predictors enter the same regression workflow as quantitative predictors.
  • Check individual coefficient evidence and residual patterns; a higher R2R^2 alone does not establish a useful predictor.

Worked Example: Household Debt and Categorical Predictors

The household-survey response is debt; existing quantitative predictors are first income X1X_1, monthly payment X2X_2, and utilities X3X_3. Their initial fitted coefficients were b0=−140,378b_0=-140{,}378, b1=0.017b_1=0.017, b2=0.96b_2=0.96, b3=157.95b_3=157.95, with R2=0.708R^2=0.708. All three slopes were reported significant.

Worked Example: Household Debt and Categorical Predictors

Adding home ownership

Let X4=1X_4=1 for owning and 00 for renting. The revised model had R2=0.712R^2=0.712 (a small increase) and standard error (18{,}594), down from (18{,}699). Its overall FF-test p-value was very small. Estimates were b0=−185,988b_0=-185{,}988, b1=0.017b_1=0.017, b2=1.05b_2=1.05, b3=200b_3=200, and b4=−12,837b_4=-12{,}837. Holding the other predictors fixed, owners were predicted to have (12{,}837) less debt than renters.

For ownership, SE(b4)=5,003SE(b_4)=5{,}003, so t=−12,837/5,003≈−2.56t=-12{,}837/5{,}003\approx-2.56, with 495 degrees of freedom; p=0.01<0.05p=0.01<0.05. Reject H0:β4=0H_0:\beta_4=0. Residual plots were described as balanced around zero, with some high standardized residuals. Ownership has evidence of association but adds little explanatory improvement, illustrating that significance and predictive gain are distinct questions.

Adding four-level location

Location has Southeast, Southwest, Northeast, and Northwest levels. Use three indicators: X5=1X_5=1 for Northwest, X6=1X_6=1 for Northeast, X7=1X_7=1 for Southwest, and all three zero for Southeast, the baseline. In addition to the prior predictors, the location model had R2=0.714R^2=0.714 and standard error (18{,}589), only a slight improvement. Estimates included b0=−184,243b_0=-184{,}243, b1=0.015b_1=0.015, b2=0.96b_2=0.96, b3=200b_3=200, b4=−12,857b_4=-12{,}857, b5=4,229b_5=4{,}229, b6=2,935b_6=2{,}935, and b7=5,211b_7=5{,}211. The location coefficients are estimated debt differences from Southeast, controlling for the other variables.

The location tests gave:

Comparison with Southeast Coefficient Standard error t=b/SEt=b/SE (p) Decision at 0.050.05
Northwest 4,229 3,471 1.221.22 0.220.22 Do not reject β5=0\beta_5=0
Northeast 2,935 2,531 1.161.16 0.250.25 Do not reject β6=0\beta_6=0
Southwest 5,211 2,881 1.811.81 0.070.07 Do not reject β7=0\beta_7=0

None of the three location coefficients met the 0.050.05 threshold. The model represents eight combinations of ownership and location, but that flexibility adds complexity. Since the three original quantitative predictors already explained substantial variation, the small gain from adding ownership and location may not justify all added parameters. Adding variables simply because they are available risks overfitting; use context, fit, individual evidence, and diagnostics together.

Key takeaways

  • Ownership was statistically significant (p=0.01p=0.01) but improved R2R^2 by only 0.0040.004.
  • Location added three comparisons to a baseline; none had p<0.05p<0.05, and R2R^2 rose only 0.0020.002 further.
  • A strong starting model leaves less unexplained variation for new predictors to capture.
  • Prefer a useful balance of explanatory performance and model simplicity over indiscriminate inclusion.

Worked Example: Flat Prices in Mysuru

For 30 flats in Jayalakshmi Puram, selling price YY is measured in lakhs, and area X1X_1 in square feet. The area-only model was Y^=28.37+0.0968X1\hat{Y}=28.37+0.0968X_1, with R2=0.44R^2=0.44. Bedroom count and bathroom count were not retained because they were highly correlated with area (multicollinearity).

Worked Example: Flat Prices in Mysuru

Premium location indicator

Let X2=1X_2=1 for a premium locality and zero otherwise. The two-predictor model improved to R2=0.66R^2=0.66, adjusted R2≈0.63R^2\approx0.63, and standard error 56.3556.35 (versus 70.5770.57). SSR=165,273SSR=165{,}273, SSE=85,728SSE=85{,}728, and SST=251,001SST=251{,}001; indeed SSR/SST≈0.658SSR/SST\approx0.658. With 2 predictors and 27 error degrees of freedom, MSR=165,273/2≈82,637MSR=165{,}273/2\approx82{,}637 and MSE=85,728/27≈3,175MSE=85{,}728/27\approx3{,}175.

F=MSRMSE=82,6373,175≈26.03F=\frac{MSR}{MSE}=\frac{82{,}637}{3{,}175}\approx26.03

with p-value close to zero for the overall test H0:β1=β2=0H_0:\beta_1=\beta_2=0. Estimates were b0=−3.92b_0=-3.92, b1=0.087b_1=0.087, b2=86.2b_2=86.2. A one-square-foot increase corresponds to an estimated 0.0870.087 lakh (₹8,700) higher price, holding location fixed. Premium location adds an estimated ₹86.2 lakh at the same area.

For area, SE(b1)=0.016SE(b_1)=0.016, so t=0.087/0.016≈5.31t=0.087/0.016\approx5.31, 27 degrees of freedom, p≈0p\approx0. For premium location, SE(b2)=20.95SE(b_2)=20.95, so t=86.2/20.95≈4.11t=86.2/20.95\approx4.11, 27 degrees of freedom, p≈0p\approx0. Both were significant. Standardized residuals were all below 3 in absolute value; plots against area, location, and fitted price were described as patternless and balanced.

Adding amenities and gated community

Now let X3=1X_3=1 for good amenities and X4=1X_4=1 for a gated community. With area and premium location retained, the model achieved R2=0.80R^2=0.80, adjusted R2=0.77R^2=0.77, and standard error 44.6444.64. SSR=201,194SSR=201{,}194, SSE=49,808SSE=49{,}808, and SST=251,001SST=251{,}001. The transcript reports MSR=50,298MSR=50{,}298, MSE=1,992MSE=1{,}992, and F=25.25F=25.25 with p-value close to zero. (The displayed MSRMSR is approximately SSR/4SSR/4; using the rounded displayed values gives an FF near 25.2525.25.)

The estimated equation is:

Y^=−43.99+0.078X1+95.66X2+35.74X3+55.38X4.\widehat{Y}=-43.99+0.078X_1+95.66X_2+35.74X_3+55.38X_4.

The spoken account gives the intercept once as −43.99-43.99 and later writes 44.9944.99; the former is consistent with the reported coefficient. For a fixed set of other predictors, one more square foot is associated with 0.0780.078 lakh (₹7,800). Premium locality adds 95.6695.66 lakhs, amenities 35.7435.74 lakhs, and gated community 55.3855.38 lakhs in the equation (the spoken interpretation rounds the last coefficient to 55.3455.34).

Area and premium locality had p-values close to zero; amenities had p=0.04p=0.04; gated community had p=0.003p=0.003. All met the 0.050.05 criterion. Adding the categorical predictors changed the area coefficient from 0.0870.087 to 0.0780.078, consistent with area alone previously absorbing some differences associated with omitted characteristics. Residual plots were described as random and standardized residuals below 3 in absolute value. This example shows that qualitative predictors can explain meaningful price variation, while bedroom/bathroom predictors need not be added when they overlap strongly with area.

Key takeaways

  • Do not add highly correlated measures of the same underlying feature automatically.
  • Premium location improved fit; location, amenities, and gated-community indicators were each significant in the expanded model.
  • Interpret every slope holding the remaining predictors fixed and preserve the units of YY and XX.
  • Validate the added fit with coefficient tests and residual diagnostics.

Worked Example: Stroke-Risk Score

A hospital data set has 25 observations. The response is a stroke-risk score from 1 to 100; predictors are patient age, blood pressure, and smoking history. Descriptive summaries were: risk mean 26.526.5, median 2626, range 77–5858, standard deviation about 1515; age mean 7070, median 7373, range 5555–8787, standard deviation about 99; blood pressure mean 156156, median 155155, range 9191–212212, standard deviation 3232. The respective coefficients of variation for risk, age, and blood pressure were about 0.560.56, 0.130.13, and 0.210.21.

Worked Example: Stroke-Risk Score

Separate and joint quantitative models

Risk plotted separately against age and blood pressure showed trends but substantial scatter; each simple trend had R2<0.3R^2<0.3. The age-only estimate was Y^=−31.75+0.832age\hat{Y}=-31.75+0.832\text{age}, with R2=0.27R^2=0.27, MSE=164.6MSE=164.6, MSR=1415.7MSR=1415.7, and F=8.6F=8.6 (p-value close to 0.010.01). The blood-pressure-only model was Y^=−6.85+0.214pressure\hat{Y}=-6.85+0.214\text{pressure}, R2=0.22R^2=0.22, MSE=176MSE=176, MSR=1156MSR=1156, and F=6.6F=6.6 (p-value close to 0.020.02). Standardized residuals in both were within ±2.5\pm2.5, with no strong residual pattern reported.

Putting both predictors in one model yielded R2=0.75R^2=0.75, adjusted R2=0.73R^2=0.73, standard error 7.697.69, hence MSE=7.692≈59.2MSE=7.69^2\approx59.2. SSR=3,899SSR=3{,}899, SSE=1,302SSE=1{,}302, and MSR=1,949MSR=1{,}949, so:

F=1,94959.2≈32.9F=\frac{1{,}949}{59.2}\approx32.9

with p-value close to zero for H0:β1=β2=0H_0:\beta_1=\beta_2=0. Both individual p-values were close to zero. The coefficients were b1=1.232b_1=1.232 for age and b2=0.334b_2=0.334 for pressure; their tt-statistics were 6.816.81 (standard error 0.180.18) and 6.476.47 (standard error 0.050.05), respectively. Both estimated effects were positive. Standardized residuals were between −2.5-2.5 and 2.52.5, and residual plots against each predictor appeared random.

The joint R2R^2 can greatly exceed either separate R2R^2: each one-predictor model leaves the other predictor out, while the joint model lets both explain response variation. Omitted predictors can distort the apparent relationship attributed to an included predictor; this is omitted-variable bias.

Adding smoking history

Let X3=1X_3=1 for a history of smoking and zero otherwise. The estimated model was:

Y^=−96.4+1.04X1+0.30X2+8.08X3\widehat{Y}=-96.4+1.04X_1+0.30X_2+8.08X_3

where X1X_1 is age and X2X_2 is blood pressure. R2=0.81R^2=0.81, adjusted R2≈0.79R^2\approx0.79, standard error 6.796.79, and MSE=46.2MSE=46.2. The reported sums of squares were SSR≈4,232SSR\approx4{,}232, SSE≈971SSE\approx971, SST=5,202SST=5{,}202. With three predictors, MSR=4,232/3≈1,410MSR=4{,}232/3\approx1{,}410; error degrees of freedom are 25−3−1=2125-3-1=21, so MSE=971/21≈46.2MSE=971/21\approx46.2.

F=1,41046.2≈30.5F=\frac{1{,}410}{46.2}\approx30.5

with p-value close to zero for the joint null that all three slopes are zero. For age, SE=0.17SE=0.17 and t=1.04/0.17≈5.95t=1.04/0.17\approx5.95, 21 degrees of freedom, p≈0p\approx0. The blood-pressure coefficient was also reported significant. For smoking, SE=3.01SE=3.01 and t=8.08/3.01≈2.68t=8.08/3.01\approx2.68, 21 degrees of freedom, with p-value reported close to zero; reject H0:β3=0H_0:\beta_3=0. At the same age and pressure, the fitted score is 8.088.08 points higher for the smoking-history group. Standardized residuals remained within ±2.5\pm2.5; residual plots against age, pressure, and smoking indicator showed both signs without a notable pattern.

Key takeaways

  • Weak one-at-a-time fit does not preclude a stronger joint model.
  • The FF-test assesses whether all slopes are zero; tt-tests assess individual slopes conditional on the other predictors.
  • Smoking adds an adjusted group difference of 8.088.08 risk-score points in this fitted model.
  • Residual checks accompany R2R^2 and significance tests; none alone establishes a complete model.

Worked Example: Monthly Online Spending

For 600 randomly selected customers, monthly website spend YY is recorded in thousands of rupees, annual salary in lakhs, married status is 11 for married, and gender is 11 for male. Mean and median spending were roughly ₹5,000, range ₹0–₹12,000, and standard deviation about ₹3,000. Salary mean and median were ₹18 lakh, range ₹12–₹25 lakh, and standard deviation about ₹4 lakh. Salary-spend correlation was 0.170.17.

Worked Example: Monthly Online Spending

Salary-only model

The fitted model was Y^=2.64+0.134salary\hat{Y}=2.64+0.134\text{salary}, with R2=0.029R^2=0.029. Since the simple-regression R2R^2 is the squared correlation, 0.172=0.0289≈0.0290.17^2=0.0289\approx0.029. ANOVA gave MSE=8.6MSE=8.6, MSR=157.3MSR=157.3, and:

F=157.3/8.6≈18.28F=157.3/8.6\approx18.28

with p-value close to zero. Thus salary was statistically significant despite the low correlation and low explained proportion; the lecture retained it for subsequent models. Standardized residuals were between −2-2 and 22, and the residual plot against salary had no strong pattern.

Add marital status

With X_2=1 for married, the reported fit was R^2=0.62, adjusted R^2 close to 0.620.62, standard error 1.831.83, MSE=3.36, and MSR=1{,}648:

F=1,648/3.36≈490.5reported as 490.8F=1{,}648/3.36\approx490.5\quad\text{reported as }490.8

with p-value close to zero. The fitted equation was:

Y^=0.84+0.10salary+4.584X2.\widehat{Y}=0.84+0.10\text{salary}+4.584X_2.

Salary had t=5.16, standard error 0.190.19, and p-value close to zero. Marital status had t=30.6, standard error 0.150.15, and p-value close to zero. Since the response is in thousands of rupees, each additional lakh in salary corresponds to ₹0.10 thousand (₹100) higher predicted monthly spending, holding marital status fixed; a ₹10-lakh salary increase corresponds to ₹1,000. Married customers have ₹4.584 thousand (about ₹4,584) higher predicted spending than unmarried customers at the same salary. Residuals against salary and marital indicator were balanced; however, some standardized residuals were above 2.52.5 or below −2.5-2.5, suggesting unexplained variation.

Gender comparison and combined model

Replacing marital status with X_2=1 for male gave R^2=0.17, adjusted R^2\approx0.164, standard error 2.722.72, MSE=7.4, MSR=442, and F=59.8 with p-value close to zero. The equation was:

Y^=1.35+0.14salary+2.2X2.\widehat{Y}=1.35+0.14\text{salary}+2.2X_2.

Both salary and gender coefficients had p-values close to zero. At a fixed salary, the model estimated male customers’ monthly spend to be ₹2.2 thousand (₹2,200) higher. Despite significance, this model explained only about 16% of spending variation; the estimated group difference’s magnitude is less firmly informative when much variation remains unexplained. Residuals looked random, with most standardized residuals between −2-2 and 22.

Including salary, married status X_2, and male indicator X_3 yielded R^2=0.76, standard error 1.471.47, MSE=2.16, MSR=1{,}339, and F=620, p-value close to zero. The transcript says adjusted R^2 was close to 0.620.62; this conflicts with its R^2=0.76 and is retained as an apparent reporting inconsistency rather than silently corrected. The estimated equation was:

Y^=−0.4+0.11salary+4.58X2+2.2X3.\widehat{Y}=-0.4+0.11\text{salary}+4.58X_2+2.2X_3.

Salary t=6.8, married t=38.1 with standard error 0.120.12, and male t=18.2 with standard error 0.120.12; all p-values were reported close to zero. At fixed values of the other predictors, each additional salary lakh predicts about ₹0.11 thousand (₹110) more monthly spend; married status predicts ₹4.58 thousand more; male status predicts ₹2.2 thousand more. Residuals against all three predictors were balanced, though some standardized residuals approached +3+3, indicating remaining variability and possible value in considering additional relevant predictors.

Key takeaways

  • A small correlation or R^2 can coexist with statistically significant predictor evidence.
  • Interpret monetary coefficients using the actual units: response in thousands, salary in lakhs.
  • Marital status improved the model much more than gender alone; the combined model included both significant indicators.
  • Significance does not imply that the model explains most individual variation or that every coefficient magnitude is estimated with high predictive accuracy.

Interaction Variables

A standard indicator model allows groups to have different intercepts but a common quantitative slope. An interaction allows the effect of one predictor to depend on another. It is often formed as the product of a continuous variable and a categorical variable.

Let X_1=1 for women and 00 for men, X_2 be work experience in months, and Y salary. Include the product X_1X_2:

Y=β0+β1X1+β2X2+β3(X1X2)+ε.Y=\beta_0+\beta_1X_1+\beta_2X_2+\beta_3(X_1X_2)+\varepsilon.

This remains a linear regression model because it is linear in the unknown coefficients βj\beta_j, although the predictors are multiplied.

For men (X_1=0): E[Y∣men,X2]=β0+β2X2.E[Y\mid\text{men},X_2]=\beta_0+\beta_2X_2.

For women (X_1=1): E[Y∣women,X2]=(β0+β1)+(β2+β3)X2.E[Y\mid\text{women},X_2]=(\beta_0+\beta_1)+(\beta_2+\beta_3)X_2.

Thus β1\beta_1 is the women-men intercept difference at X2=0X_2=0, β2\beta_2 is the men's experience slope, and β3\beta_3 is the difference in experience slopes. The gender indicator is a moderator because the salary-work-experience relationship depends on gender.

Interaction Variables

Worked salary example

For a 30-person sample, the fitted model was:

Y^=13,450.2−7,771.68X1+293.8X2−241.6X1X2.\widehat{Y}=13{,}450.2-7{,}771.68X_1+293.8X_2-241.6X_1X_2.

The model had R^2=0.92, adjusted R^2=0.91, overall F-test p-value small, and each predictor p-value below 0.050.05.

For women, substitute X1=1X_1=1:

\widehat{Y}_{\text{women}}&=13{,}450.2-7{,}771.68+(293.8-241.6)X_2\\ &=5{,}678.52+52.2X_2\approx5{,}678.5+52.2X_2. \end{aligned}$$ For men, substitute $X_1=0$: $$\widehat{Y}_{\text{men}}=13{,}450.2+293.8X_2.$$ The fitted salary increase per extra month of experience is $52.2$ for women and $293.8$ for men. These are conditional slopes; the interaction coefficient $-241.6$ captures the difference between them. **Key takeaways** - Without interaction, groups share a slope; with interaction, the slope can vary by group. - The coefficient on the product term is the difference in slopes for a binary moderator. - A model can be nonlinear in predictors yet linear in its coefficients. - Substitute each indicator value to derive and interpret group-specific equations. #### Module Synthesis: Model Building and Inference Categorical variables extend multiple regression to mixed quantitative and qualitative data. Least squares estimation, multiple $R^2$, adjusted $R^2$, the overall $F$-test, individual $t$-tests, and residual analysis remain central. In a sample, these procedures support inferences about population relationships under the regression assumptions. Model building is iterative: propose predictors with substantive reason, estimate the model, interpret effects with units and comparisons explicit, assess tests and residuals, and decide whether improved explanation justifies added complexity. A large $R^2$ does not make every predictor useful; a statistically significant coefficient can coexist with low explanatory power. Correlated predictors can obscure individual contributions, while omitted relevant variables can distort coefficients. Qualitative data can add substantial explanatory value, as in the flat-price and spending examples, but other categories added little in the household-debt example. The broader purpose of business statistics is to summarize data, analyze it, make evidence-based inferences for decisions beyond the observed sample, and quantify uncertainty. This module’s regression tools build on descriptive statistics, probability, sampling distributions, confidence intervals, hypothesis tests, and simple and multiple regression. Statistical software reduces calculation effort; sound interpretation and decisions remain the analyst’s responsibility. **Key takeaways** - Use indicators for categories and interaction terms when a slope may depend on group membership. - Assess overall fit, individual predictor evidence, and residual behavior together. - Keep only predictors that contribute useful, defensible information; do not add variables merely because they were collected. - Regression supports population inference from sample data when its assumptions are appropriate.

Test yourself on this module

Questions from these notes, one at a time, with feedback. Your weak topics are saved.

Set it up free Takes about a minute. Free plans work.